ES2279749T3 - TIMING SYNCHRONIZATION PROCEDURE OF A DIGITAL SIGNAL. - Google Patents

Abstract

Synchronization involves an input analogue signal (r(t)) being applied to an adaptive filter (52,53) of samples (r(mTe). The adaptive filters provide correlation samples (y(m),z(m)). The elementary power of each sample is calculated, and a sliding window defined a set size larger than the sample period. The window is swept across a field at an instant of time. For each window the elementary power of the correlation sample are calculated, and the maximum power window determined. This maximum determines the synchronization window position.

Description

One-beat rhythm synchronization procedure digital signal

Technical field

The present invention aims at a rhythm synchronization procedure of a digital signal. Find an application in transmission systems radio, and more particularly in systems with access multiple by code division (AMDC in abbreviation or "CDMA" in English for "Code Distribution Multiple Access").

Prior art

The principles of a communication are known digital and the connection between baseband signals and signals on carrier frequency and are described, for example, in the work of John G. PROAKIS entitled "Digital Communications", McGraw Hill International Editions.

The principle scheme of a chain of Digital radio transmission is provided in Figure 1.

In the emission chain E, a digital signal 10A of origin that wants to transmit experiences a previous treatment in a circuit 10. This pretreatment may comprise various randomization, interleaving or coding operations, which do not They will be discussed below in the description. Circuit 10 supplies a 10B sequence of digital symbols, called a (k), in which k designates the range of the symbol. The symbol a (k) is in general a complex number, represented by a pair of real values. The frequency of the symbols a (k) is denominate Hs and the corresponding period is called Ts, being Ts = 1 / Hs.

From the sequence a (k), a editing device 20 produces analog band signal 20A base to be issued, called b (t), in which t designates The time variable. The signal b (t) is a complex signal represented by two signals b_ {I} (t) and b_ {Q} (t) real quadrature. The baseband signal 20A is transposed over a carrier frequency by a radio transmitter 30. This issuer It comprises different devices: modulator, transpositions of frequency, filters, local oscillators, amplifiers, antenna, which will not be discussed next. Only the issuer is supposed to performs a linear mathematical operation with respect to the signal of base band The emitted radio signal 30A is propagated to the receiver while experiencing various kinds of degradations.

In the receiving R chain, the signal 40A radio received is first treated by a receiver  40 radio This receiver comprises different devices: antenna, frequency transposition media, filters, oscillators premises, amplifiers, which will not be discussed below. He receiver 40 supplies a baseband analog signal 40B, called r (t). The signal r (t) is a complex signal, represented by two signals r_ {I} (t) and r_ {Q} (t) real quadrature. From the signal r (t), a detection device 50 produces a series of symbols or 50A digital samples. The series of samples detected 50A it constitutes a more or less faithful image of the sequence of symbols a (k). 50A detected samples experience a post treatment in a circuit 60. This post treatment It includes various operations that correspond to the operations of pretreatment 10 of the emission chain E and supplies the signal restored 60A.

The elaboration of the series of detected samples 50A assumes that exactly the value of the period Ts of the sequence rhythm a (k) and its phase with respect to the baseband signal r (t) is known. A synchronization device 70, thanks to the signals 50B that it exchanges with the detection device 50, estimates the rhythm of the received signals and communicates the result of this estimation to the detection device. Certain detection procedures, called coherent demodulation, also require knowledge of the phase of the carrier frequency of the received radio signals. This knowledge is not considered in this document, and demodulation can be both coherent and non-coherent. It is only assumed that the beat frequency between the carrier frequency
used in the receiver and the actual carrier frequency is low compared to the rhythm Hs of the symbols a (k).

The functional cut that has just been made it presents a certain arbitrariness, and certain operations can be complicated However, the effective presence of the 40B baseband signal received or from a representation equivalent of this signal in the form of digital samples.

The present invention essentially relates to the synchronization operation implemented in the chain of reception.

The synchronization is related to the editing of the baseband signal b (t) to be broadcast and with the corresponding detection conditions. In the present document the edition is supposed to correspond to the operation mathematically linear:

b (t) = \ sum \ limits_ {k} a (k) \ cdot h (t-kTs)

where \ sum \ limits_ {k} represents a sum over all symbols a (k) and in which h (t) designates a real or complex function of time t.

An important case is that of widening by direct sequence in which h (t) = \ sum \ limits_ {n = 0} N-1} \ alpha (n) \ cdotg (t-nTc).

In this expression, α (n) is a family of real or complex numbers previously defined e independent of the value of the symbols a (k) they want be transmitted The numbers α (n) are called "chips" according to a terminology widely used in this technique. N chips? (N) are associated with each rank k successive numbered from n = 0 to n = N-1. The chips are supplied with a period Tc = Ts / N and the corresponding rhythm is called Hc. The number N of chips per symbol is called the widening The function g (t) is a real function or complex independent of the range k and the number n. It is called "editing function" of the chip. AMDC systems of direct sequence spread transmission attributed to each user a particular family of α (n) chips, selecting the different chip families so that Reduce interference between users.

Although the invention does not refer directly to the detection itself, it is necessary to take it into account. In the case of a transmission channel that does not deform the signal but that only an independent disturbing signal overlays called Gaussian white noise, the optimal detection is obtained by the adapted filtration method or an equivalent method. This method consists in applying the baseband signal r (t) received to a transfer function filter h * (- t) for get a signal s (t):

s (t) = h \ text {*} (- t) * r (t)

in which the * sign represents the complex conjugation operation when placed as an exponent, or the convolution operation when placed at mid-height. With total rigor, the adapted filtration is done with the help of a transfer function filter h * (Tr-t), in the that Tr is a fixed delay selected so that the function h * (Tr-t) is causal with respect to the variable t. This delay corresponds to the time required to finish the calculation of s (t) from r (t) for an instant t given, but does not play any role in the explanations that follow. For simplicity, it is assumed void below. The values of the signal s (t) in moments conveniently selected constitute the series of samples detected 50A or will allow, with the help of complementary operations, to prepare this Serie.

In the case of broadening by sequence direct, the adapted filtration decomposes in a filtration adapted to the shape of the chip:

s_ {c} (t) = g \ text {*} (- t) * r (t)

and a filtration adapted to the sequence of chips:

s (t) = \ sum \ limits_ {n = 0} N-1} \ alpha \ text {*} (n) \ cdot s_ {c} (t + nTc)

in which the function s_ {c} (t) is used as a calculation intermediary. The filtration adapted to the chip sequence is called widening.

There are numerous recovery procedures of the rhythm from the baseband output signal of the filter adapted. Some resort to a global approach in which the phase of the carrier frequency, the rhythm phase and the symbols are jointly estimate. From a practical point of view, it is often easier to estimate the rhythm phase separately. By way of In general, rhythm recovery needs a shunt with with respect to the time of the filter baseband output signal adapted, in order to highlight the transmissions of the signal. Among the possible procedures, some gather a differentiator and a phase lock loop, others perform a non-linear operation followed by filtration and a step detector at zero of the signal, others multiply the signal by itself delayed

Actually, the radio signal is propagated often in a complex way between the sender and the receiver following Several different paths. It is the case of the mobile radio channel land. The received signal is presented at the receiver in outdated moments. The optimal detection should take into account the impulse response of the channel. From the point of view mathematical, the filtration adapted to the impulse response of the channel performs a recombination of all existing paths. This operation is partially performed on classic receivers. rake (called "rake receiver"), which combine a unlimited number of journeys The number of devices of Elemental pace recovery is as high as the number of propagation paths treated.

The patent US-A-5 778 022 describes a rhythm synchronization procedure of a digital signal according to the preamble of claim 1.

The objective of the invention is to implement a common rhythm recovery procedure that you can do against a very high number of journeys.

Description of the invention

In the case of an ideal transmission without noise, the received signal r (t) of baseband is related to the baseband signal b (t) to be emitted by an expression Shape:

r (t) = A \ cdot exp (j \ varphi) \ cdot b (t- \ tau )

in which j designates the part imaginary of a complex number (j2 = -1) and exp the function exponentially, A and var represent respectively the amplitude and the phase of the gain of the transmission channel and? the time of propagation. If the carrier frequency is not known exactly, the phase term? varies slowly with the course weather; this term is canceled or compensated in the case of a consistent demodulation. The output signal s (t) of the filter adapted is from the shape:

s (t) = A \ cdot exp (j \ varphi) \ cdot \ sum \ limits_ {k} a (k) \ cdot Rh (t- \ tau - kTs)

in which Rh (t) designates the temporal autocorrelation function of h (t). In general it select the function h (t) so that its correlation Rh (t) verifies the said condition of Nyquist:

Rh (nTs) = 0

for any integer n nonzero, and takes low values when the variable t takes values greater than some periods of the symbol Ts. In these conditions, the value of s (t) at the moment \ tau + nTs is:

s (\ tau + nTs) = A \ cdot exp (j \ varphi) \ cdot a (n) \ cdot Rh (O)

and its value in an instant \ theta + \ tau + nTs, in which the quantity q is small with respect to to Ts, it's from approximately:

s (\ theta + \ tau + nTs) \ cong A \ cdot exp (j \ varphi) \ cdot a (n) \ cdot Rh (\ theta)

The square of the quantity module s (\ theta + \ tau + nTs) is:

| s (\ theta + \ tau + nTs) | 2 \ cong A 2 \ cdot | to (n) | 2 \ cdot | Rh (\ theta) | 2

If we are interested in the case of a digital modulation by phase shift, the value of | a (n) | ^ {2}
it is then independent of the transmitted symbol and the quantity | s (the + \ + nTs) | 2 is approximately proportional to
| Rh (the) | 2. The autocorrelation function Rh (t) takes its maximum value at t = 0. If the function h (t) is properly selected, the correlation function Rh (t) has a narrow main maximum in the vicinity of t = 0 and eventually secondary highs very attenuated when the value of t increases. The observation of the position of the successive maximums of | s (t) | 2 then allows us to locate the instants \ + nTs and thus recover the rhythm. Next, the quantity | s (t) | 2 will be referred to as signal strength after the adapted filtration.

In the case of a multiple transmission paths, there is no single propagation time and, with evolution of the power | s (t) | 2 over the course of time, successive maximum packages located in the proximities of the middle position of each symbol a (k). Every maximum correlation corresponds to a propagation path particular and each maximum package occupies a time interval which corresponds to the difference between the propagation time plus Long and shorter time. It is called "path window"  at such a time interval. The procedure of synchronization proposed by the invention consists in placing a window on the maximum correlation packages. The center of the window then defines an average propagation time \ tau and the position of the different paths is located with respect to This weather.

The received r (t) baseband signal supplied by the radio receiver has experienced a low pass filtration. Therefore its frequency spectrum has an upper limit The same is true for the h (t) function of edition. As a consequence, the signal s (t) also has a frequency limit From the mathematical point of view, if Fmax designates the highest frequency contained in the spectra of the r (t), h (t) and s (t) functions, these functions are they represent exactly by their samples r (mTe), h (mTe), s (mTe) taken in moments mTe regularly separated by a quantity Te, called the sampling period, to condition that this amount verify the called condition of Shannon:

You < \ frac {1} {2 \ cdot F max}

       \ newpage
    

For convenience, it can be selected here as Sampling period Te a submultiple of the symbol period Ts:

Te = \ frac {Ts} {M}

in which M is an integer greater than 1 and compatible with Shannon's condition. It is not indispensable to select Te as a submultiple of Ts, but it is indispensable that the selection of M be compatible with the condition from Shannon. Samples of s (t) are expressed with the help of samples of r (t) and h (t):

s (mTe) = Te \ cdot \ sum \ limits_ {i} h \ text {*} (iTe) \ cdot r [(i + m) \ cdot Tea]

For convenience, work with the quantity:

and (m) = s (mTe) / Te

since the factor Te does not play no paper From a practical point of view, the function h (t) takes low values from a certain value of t and the sum over i can be limited to an interval between two relative numbers i_ {1} and relative i_ {2} selected in principle according to the decrease of h (t):

y (m) = \ sum \ limits_ {i = i_ {1}} ^ {i_ {2}} h \ text {*} (iTe) \ cdot r [(i + m) Te]

Samples and (m_ {0}), y (m_ {0} +1), ..., and (m_ {0} + Ne-1) successive, in which m_ {0} is a particular range and Ne a number given positive integer, they are placed in an initial instant window m_ {0} Te and of length NeTe. The integer Ne is selected, by Of course, depending on the dispersion of the paths. Containing a period Ts of symbol M sampling periods Te, can represent the sampling range m with the form:

m = M \ cdot p + q

in which p and q are whole numbers relative. For a given integer q_ {0}, when q takes the values from q_ {0} to q_ {0} + Ne-1 and p takes all possible integer values, you get a window of NeTe length that reproduces with the period Ts.

The synchronization procedure of the invention is based, on the one hand, on the calculation of total power in the window:

\ sum \ limits_ {m = m_ {0}} ^ {m_ {0} + Ne-1} | y (m) | ^ {2}

And, on the other hand, in the distribution of powers | and (m) | 2 inside the window.

Synchronization comprises two modes: one mode Acquisition and a tracking mode. In acquisition mode, the M possible positions, for q = 0 in M-1, of the periodic window [Mp + q; Mp + q + Ne-1] are examined successively and the total power, called P (p, q) in each window:

P (p, q) = \ sum \ limits_ {i = 0} ^ {Ne-1} | y (Mp + q + i) | 2

This power is higher when there are paths in the window that when there are none. At the moment when the total power takes its maximum value, the range that has a certain value q_ {0} and the window stops about the package of paths.

Then it goes into tracking mode. To each range i inside the window a weight is associated c (i) depending on its position with respect to the center of the window. This weight is a monotonous function (in the broad sense) of rank i. The sum of the weighted powers:

\ sum \ limits_ {i = 0} ^ {Ne-1} c (i) \ cdot | y (Mp + q_ {0} + i) | 2

indicates where the power is located average of the package of the routes with respect to the center of the window and can be used to correct the sampling time and thus feed back the position of the window on the package.

Precisely, the present invention has therefore as an object a rhythm synchronization procedure of a digital signal, in which it is sampled with a certain period of sampling an analog signal from the transmission of a signal modulated with the help of an editing function, a adapted filtration of the samples, this filtration being adapted to the editing function used by the modulation and that leads to correlation samples, characterizing this procedure because:

- the elementary power of each is calculated correlation sample;

- a sliding window of length is defined Ne times the sampling period, that is, NeTe, and that begins in a certain range;

- for each sliding window, the sum of the elementary powers of the correlation samples located in this window for a symbol or for a number set of symbols;

- the window for which the sum is determined of powers is maximum;

- synchronization is then defined by the position of the synchronized window on the window in which the sum of powers is maximum, and by the range of each sample of correlation inside this window.

Preferably, two types are implemented of operations and proceed in two modes:

a) in a first type of operations, called exploration operations, all the following are examined successively Possible positions of the sliding window (called cycle of scan), and for each position, the overall power is calculated (Pa) of the correlation samples (and (m)) contained in the sliding window, the window for which the power is identified global is superior from the beginning of the cycle to the position current, and the value of this higher power (Pam) is memorized,

b) in a second type of operations, called follow-up operations, only the correlation samples (z (m)) in which the range is in a window called tracking window, is calculated on the one hand the overall power (Pb) of these samples, on the other hand a signal (d) that allows to feed back the center of this window on the average position of the elementary powers it contains,

c) in a first mode of operation, called acquisition mode,

-

by a part, every time a sliding window appears a overall power (Pa) greater than the last memorized power (Pam) from the beginning of the scan cycle to the current position, the current position of the sliding window and a procedure of check,

-

by another part, when the scan cycle is finished, it goes to a called tracking mode,

d) in a second mode of operation, called of follow up,

-

by one part, the transfer mechanism of the position of the sliding window to the tracking window,

-

by otherwise, when permanent verification fails, it returns to the acquisition mode

Brief description of the drawings

- Figure 1, already described, illustrates the principle of a digital transmission chain;

- Figure 2 is a scheme of detection and of the synchronization implemented according to the invention;

- Figure 3 is a functional scheme of the sequencing;

- Figure 4 shows various schedules that they illustrate sampling, sliding windows, powers in the windows and tracking and editing ranges;

- Figure 5 illustrates the distribution of powers in exploration and monitoring;

- Figure 6 illustrates a particular way of realization of de-widening by a method of records of displacement;

- Figure 7 illustrates another embodiment of the widening by a method of accumulation;

- Figure 8 shows the operation of adapted and synchronization filtration in the particular case of a AMDC-MDP2 transmission.

Detailed description of particular embodiments

Before describing certain particular modes of performing the procedure, it is useful to define the different phases of the procedure:

Scanning operations

It's about window scrolling sliding, of the power calculations in this window, of the search for the position that provides maximum power.

Tracking operations

It's about the power calculations in the tracking window (even if you have not yet obtained the synchronization), of the calculation of the deviation signal of the position of this window, of the feedback of the rhythm of Sampling by this deviation signal.

Acquisition mode

In this mode, in the course of a cycle of scan, every time a power in the sliding window is greater than the highest power recorded since the beginning of the cycle, the position of the sliding window is communicated to the window of tracing. In general the synchronization is obtained before the end of the exploration cycle, but only ensures that you have gone through the maximum power position at the end of the cycle. So It goes into tracking mode.

Tracking mode

In this mode, the tracking window is considered to be synchronized. Therefore, any action of the scanning device on the position of the tracking window is prohibited. The permanent verification procedure is the one that allows to discover an eventual desynchronization and decide to return to the acquisition mode
tion.

check

In acquisition mode, the transfer of position of the sliding window to the tracking window triggers a verification procedure. This procedure re-launch on each new transfer. It is not interrupted by the transition from acquisition mode to tracking mode.

Synchronization with material savings

In this case, a single set of adapted filtration that, in acquisition mode, performs the scanning operations, then in tracking mode, performs Tracking operations Therefore it is necessary to wait for end of the scan cycle to transfer the position of the sliding window of maximum power to the window tracing.

Figure 2 generally illustrates the detection and synchronization operations. It is a scheme functional in which the clipping does not necessarily correspond to a material trimming and in which only the necessary elements appear for the understanding of the invention. The detection device 50 It is composed of a sampling device 51, two sets 52 and 53 of adapted filtration, and by a set 54 of demodulation and treatment of propagation paths. He synchronization device 70 is composed of a generator 71 of rhythm, a scanning device 72, a device 73 of monitoring and a sequencing system 74.

The rhythm generator 71 supplies a signal Periodic 71A, called sampling and designated by He, of period Tea. From the signal He and the signal received r (t) from baseband, the sampling device 51 develops the 51A series of r samples (mTe). The sampling performed is asynchronous with with respect to the symbol period Ts since the sampling phase is in principle any with respect to the beginning of a period and the equality Ts = M \ cdotTe is only strictly verified when synchronization is acquired.

When driving the 72A signals sent by the scanning device 72, the adapted filtration assembly 52 performs the calculation of the samples and (m) according to the relationship already mentioned:

and (m) = \ sum \ limits_ {i = i_ {1}} ^ {i_ {2}} h \ text {*} (iTe) \ cdot r [i + m) \ cdot Tea]

It is a classic transversal digital filtration. This filtration can be obtained by sliding correlation: the samples r (mTe) travel in a shift register with the He rhythm; The fixed coefficient h * (iTe) that is used to calculate the product h * (iTe) \ cdotr [(i + m) \ cdotTe] is associated with each position i of the register; In each sampling period, a new sample and (m) are provided. Numerous variants are possible. For example, the series of coefficients h * (iTe) parades in front of the current r (mTe) sample or the samples y (m) are calculated in parallel,
y (m + M), and (m + 2M), ... corresponding to successive symbols.

From the samples and (m), the scanning device 72 calculates the power contained in the NeTe length window, that is, from i = 0 to i = Ne-1:

P (p, q) = \ sum \ limits_ {i = 0} ^ {Ne-1} | y (Mp + q + i) | 2

The transmission channel is not perfect: it contains Noise and multiple propagation paths are not stable. By therefore, it will often be necessary to take an average of the power P (p, q) on several symbols, which is equivalent to calculating the power called Pa (p, q):

Pa (p, q) = \ sum \ limits_ {w = 0} ^ {Ns-1} P (p + w, q)

in which Ns is the number of symbols considered.

When driving the 73A signals sent by the Tracking device 73 and the signals sent by the set 52 filtration adapted in exploration, set 53 of adapted filtration in follow-up performs the calculation of a series of samples z (m) according to the same relation as for samples and (m):

z (m) = \ sum \ limits_ {i = i_ {1}} ^ {1 ^ {2}} h \ text {*} (iTe) \ cdot r [(i + m) \ cdot Tea]

The difference is because the samples and (m) are calculated for the possible M relative positions of a sliding window, while the samples z (m) are calculated only for the Ne positions located inside of the synchronized window.

From samples z (m), the Tracking device 73 calculates the power, called Pu, contained in the synchronized window and the power called Pb added on several symbols:

Pu (p, q) = \ sum \ limits_ {i = 0} ^ {Ne-1} | z (Mp + q + i) | 2

Pb (p, q) = \ sum \ limits_ {w = 0} ^ {Ns-1} Pu (p + w, q)

Since the calculations are made from the z (m) samples located in the synchronized window, the values obtained for Pu and Pb can be differentiated from the values obtained for P and Pa.

Sequencer 74 directs the stages of the synchronization of the tracking device 73 from scanning device 72. Figure 3 illustrates the procedure of synchronization, by means of two state diagrams: general sequencing 80 and verification sequencing 90. In these diagrams, the rectangles represent situations particular sequencing, the arrows indicate the direction of the evolution and the text on the arrows remember the evolution conditions.

The general sequencing block 80 comprises an initialization block 82, a block 84 corresponding to a acquisition mode and a block 86 corresponding to a mode of tracing.

Verification sequencing block 90 it comprises a block 92 of idle state, a block 94 of initial verification and a block 96 of permanent verification.

After an initialization phase (block 82) to allow the different circuits of the detection and synchronization devices 50 to function correctly, the sequencer enters the acquisition mode (block 84) and performs a scan cycle of the M relative positions of the series of coefficients h * (iTe) and the series of samples
r (mTe). Each time a new value of the power Pa is greater than the previous ones, this value is memorized and the position of the sliding window is transferred to the window of the tracking device. Thus, at the end of the scan cycle, the maximum value of the power Pa is stored and the tracking device is stopped on the window of maximum power.

Taking into account the imperfections of the channel of transmission, it is not excluded that the power Pa can take momentarily an abnormally high value and that a false acquisition. As soon as a transfer of position of the sliding sliding window to the window of tracking, the power Pb obtained in the tracking window is compare with the power Pa obtained in the scanning window. From the practical point of view, the following can be adopted Verification strategy in two stages called, respectively, "initial verification" and "verification permanent ", illustrated by block 90 of Figure 3. The initial verification (block 94) determines whether synchronization proposed by the scanning device is plausible. The permanent verification aims to discover an eventual desynchronization For initial verification you can define a relative power threshold, named λ, real number between 0 and 1. Designating by Pam the power value Pa memorized at the time of a particular transfer of the window position, it is desired to compare the power Pb obtained in tracking with the value \ lambda \ cdotPam. The verification initial consists of examining between Na successive values of the Pb power if at least Nb among them are greater than Amount \ lambda \ cdotPam. The integers Na and Nb are Select according to the transmission conditions. They take Small enough for the verification time to be short (for example Na = 4 and Nb = 2). If the exam is satisfactory, it passes to permanent verification (block 96). This consists of compare the power Pb with a quantity \ mu \ cdotPo, in which Po designates the expected average Pb power in the synchronized state, and µ designates a threshold, close to 1, selected with greater precision than the threshold λ. But the comparison is made over a large number Nv of successive values of Pb in order to reduce the influence of fluctuations: for example, they can add the Pb- \ mu \ cdotPo deviations obtained  on Nv consecutive tracking windows and can be considered synchronization as satisfactory if the sum is positive, then Restart the operation with the following Nv values and repeat So. It should be noted that the initial verification and verification permanent can be performed without interruption of the procedure exploration.

In acquisition mode (block 84), the appearance of a Pa power exceeding the last memorized Pam power interrupt any verification procedure eventually in course and launches a new initial verification (block 94). In the end of the scan cycle, the sequencer goes from mode 84 of acquisition to 86 tracking mode and remains in it while not a verification failure is manifested and returned to idle status sequencing verification. Then retakes general sequencing 80 from its state of initialization

When the sequencer is in 86 mode of monitoring, filtration set 52 adapted to scanning and The scanning device 72 is made available. By way of optional, the scanning device can continue searching for what sliding sliding window position provides the maximum power Then this position can be saved at the end of Each scan cycle. In case of loss of synchronization, this saved position can be used to launch a new fast acquisition

Given that the sets of adapted filtration in exploration and follow-up perform the same type of calculations, a single set of adapted filtration that works primarily in exploration and Then follow up. With respect to the previous sequencing, only it is necessary to wait at the end of the scan cycle to undertake verification, and the average acquisition time is more long. However, the volume of calculations made in the exploration filtration set is often of an order of magnitude greater than the corresponding volume of the calculations of the filtration set on track so that the suppression of a filtration set does not provide material savings significant.

The synchronization, obtained according to the sequencing just described, is kept feedback the sampling rate generator 71 (figure 2) on the deviation of the tracking window with respect to the position considered ideal. The deviation signal, called d, is obtained from the output samples z (m) of the filtration assembly 53 in tracking as a sum of the weighted powers in the window, as mentioned above:

d = \ sum \ limits_ {i = 0} ^ {Ne-1} c (i) \ cdot | z (Mp + q_ {0} + i) | 2

in which q_ {0} s the range transmitted to the filtration assembly 53 in follow-up during the acquisition.

There are different possible choices for weights c (i). For example, c (i) = 1 can be taken in the upper half of the window and c (i) = - 1 in the lower half, which is equivalent to taking the difference in powers as a signal in each half-window. It can also be selected as the ideal position from the window the barycenter of the powers, that is, select weights c (i) of the form:

c (i) = i- \ frac {Ne-1} {2}

in which \ frac {Ne-1} {2} characterizes the center of the window. This last selection will generally lead to a synchronization more stable than the previous one. Of course also intermediate selections are possible, in which the weight c (i) it grows, in absolute value, when the range i deviates from the center of the window.

The rhythm generator 71 can be fed back, for example, with the help of a second phase lock loop order that uses the d signal as an error signal.

From the sample series z (m) of adapted filtration in follow-up and indications 73B of Hs rate of symbol supplied by the tracking device (for example, in the form of the first sampling range of the window follow-up), set 54 of demodulation and treatment of trajectories elaborates the series of symbols or samples detected 50A. This set 54 will generally perform an elementary demodulation on each window path and will recombine the results.

For convenience, in all of the above, it has been described the synchronization in the case where the function h (t) editing does not depend on the k range of the symbol. In certain important cases, the edition depends on the k rank of the symbol, but not of the symbol a (k) itself. If the functions of Successive editions are conveniently selected, they can give as result certain advantages: possibility of facing paths propagation multiples that disperse over a length exceeding the symbol period, reduce or standardize the effects of interference.

Here we consider the case of a sequence of K successive editing functions, called h (k, t) for de k = 0 to K-1, which are repeated periodically every K Symbols So synchronization is not just about locating the samples in a symbol period but also the symbols received with respect to the sequence of functions h (k, t). Then the output signal s (t) of the filter adapted must be replaced by K signals s (k, t). Set Filtration adapted in exploration supplies K series of samples y (k, m) for from k = 0 to K-1:

and (k, m) = \ sum \ limits_ {i = i_ {1}} ^ {i_ {2}} h \ text {*} (k, iTe) \ cdot r [(i + m) \ cdot Tea]

It is assumed that the functions h (k, t) of edition are conveniently selected so that their functions of intercorrelation take low values. For a given value of k, the h * (k, iTe) \ cdotr [(i + m) Te] products that have a important contribution in the calculation of y (k, m) are then those that correspond to the passage of the range symbols k + wK, in which w designates any relative integer. The power of the samples and (k, m) in a sliding window you can still used to search for synchronization.

When the number K exceeds some units, it it becomes difficult to use K adapted filters that work simultaneously. Then the following strategy can be adopted. It is assumed that the filter adapted for scanning is composed of a length shift register i_ {2} -i_ {1} -1, in which the samples r (mTe) pass through, and for a record of length i_ {2} -i_ {1} +1 containing the coefficients h * (k, iTe). In the instant (i_ {2} + m_ {0}) \ cdotTe, in the that m_ {0} is a particular integer, the record of displacement contains the samples r [(i_ {1} + m_ {0}) \ cdotTe], r [(i_ {1} + m_ {0} +1) \ cdotTe], ..., r [(i_ {2} + m_ {0}) \ cdotTe] and the registration of the coefficients with the values h * (k_ {0}, iTe) for an integer k_ {0} particular. During a symbol period, the filter adapted supplies the samples and (k_ {0}, m_ {0}), y (k_ {0}, m_ {0} +1), ..., and (k_ {0}, m_ {0} + w), ..., and (k_ {0}, m_ {0} + M-1) successive. Then load the register of the coefficients with the coefficients h (k_ {0} + 1, iTe). The adapted filter supplies the samples and (k_ {0} + 1, m_ {0} + M), ..., and (k_ {0} + 1, m_ {0} + M + w), ..., and (k_ {0} + 1, m_ {0} + 2M-1). Then you follow the operation with the coefficients h * (k_ {0} + 2, iTe) and repeat until the coefficient set h * (k_ {0} + Ns-1, iTe) for which the filter adapted provide the samples and (k_ {0} + Ns-1, m_ {0} + Ns \ cdotM-Ns), ... and (k_ {0} + Ns-1, m_ {0} + Ns \ cdotM-Ns + w),  ... and (k_ {0} + Ns-1, m_ {0} + Ns \ cdotM-1). The exam is thus performed on Ns consecutive symbols. The Ns samples y (k_ {0}, m_ {0}), and (k_ {0} + 1, m_ {0} + M), ... and (k_ {0} + Ns-1, m_ {0} + Ns \ cdotM-Ns) correspond to a given relative position of the sequence r (mTe) with respect to editing functions. Let p_ {0} be designation of this position. The samples y (k_ {0}, m_ {0} + w), and (k_ {0} + 1, m_ {0} + M + w), ..., y (k_ {0} + Ns-1, m_ {0} + Ns \ cdotM-Ns + w)  correspond to a relative position, displaced w periods of sampling with respect to the previous relative position. The designation of this position is p_ {0} + w. The power corresponding to this position is:

\ sum \ limits_ {i = 0} ^ {Ns-1} | y (k_ {0} + i, m_ {0} + iM + w) | 2

The powers corresponding to the M positions p_ {0}, p_ {0} +1, ..., p_ {0} + M-1. To get the M positions following, you must follow the operation of loading the registry of coefficients, not with the coefficients h * (k_ {0} + Ns, iTe), since they would get the positions p_ {0} a again p_ {0} + M-1, but with the coefficients h * (k_ {0} + Ns-1, iTe), that is the game of coefficients used for the examination of the previous symbol.

In what has just been exposed, it is evident, on the one hand, that the range k contained in h (k, t) e and (k, m) should be evaluated as a module of K, on the other hand that the elementary powers | y (k, m) | 2 must undergo appropriate memorization in order to be used in The convenient time in the calculations.

The examination of the possible relative positions K, as just described, may take quite a long time. To reduce this time, Ns associated filters can be used as follows. If these filters are numbered from 0 to Ns-1, the entry position of the Ns-1 filter offset register is connected to the sample
r (mTe) current; The M-th position of the shift register of each filter (except filter 0) is connected to the entry position of the offset register of the previous filter. The registration of the coefficients of filter 0 is loaded with the coefficients h * (k_ {0}, iTe), in which k_ {0} is arbitrary (for example, k_ {0} = 0), and each filter is loaded with the coefficients h * (k_ {0} + w, iTe) corresponding to their number w. In each sampling period, the Ns outputs of these filters provide Ns samples corresponding to the same relative position of the received r (mTe) sequence and the sequence of the editing functions. After K \ cdotM sampling periods, all relative positions have been examined, without changing the coefficients.

       \ global \ parskip0.950000 \ baselineskip
    

The filter adapted in tracking can also be composed of a shift register in which the samples r (mTe) and a register containing the coefficients h * (k, iTe). When the scanning device transfer the position of the window to the tracking device, communicates the k range of the editing function with which you must Start working the filter adapted to follow up. Register of the coefficients of the filter adapted in monitoring is loaded with the corresponding coefficients h * (k, iTe). Then, in each symbol period, the contents of the registry of the coefficients with the sequence of coefficients h * (k, iTe) in which the range k is increased by one unit (module K). Unlike the filtration set adapted to exploration, the complexity of filtration set adapted to follow-up is not affected by the number K. Its complexity depends only on the length Ne \ cdotTe of the tracking window. If the number of samples Ne in the window is at least twice smaller than the number of M samples in a symbol period, the length can be reduced of the filter shift register adapted to track and Sample z (k, m) in several stages. For him Otherwise, if the number Ne is greater than M, it is necessary to use Multiple filters to simultaneously obtain samples corresponding to different symbols.

The invention just described applies, in particular, by sequential widening communications direct. In that case, the editing functions h (k, t) are of the shape:

h (k, t) = \ sum \ limits_ {n = 0} ^ N-1} \ alpha (k, n) \ cdot g (t - nTc)

This expression generalizes the expression of h (t) provided above. Chip family α (k, n) is composed of K successive sequences (of k = 0 to K-1) of N chips a (k, n). Often results from the combination of a broadening sequence proper of N chips and a sequence, called Randomization of K \ N chips. Here is not considered the way of producing a randomized widened sequence of this type and not It plays no role in the invention. Only the K \ cdotN family of chips has autocorrelation properties convenient. The g (t) editing function of the Chip is common to all chips.

As indicated above, filtration adapted breaks down into a filtration adapted to the shape of the chip and a spreading by the chip sequence. Filtration adapted to the shape of the chip is common to scanning filtration and to the filtration adapted in follow-up. It is supplied in the mTe instants of samples x (m):

x (m) = \ sum \ limits_ {i} g \ text {*} (iTe) \ cdot r [(i + m) \ cdot Tea]

Therefore the sample series is obtained x (m) by a classical digital filtration in the rhythm of sampling He. To comfortably perform the widening by chip sequence, it is necessary to take the sampling period as submultiple of the chip period:

Te = \ frac {Tc} {Nc}

in which Nc is an integer greater than 1. Then the number M of samples per symbol is: M = N \ cdotNc.

From the series of samples x (m), the flattening supplies the sample sequences and (k, m):

and (k, m) = \ sum \ limits_ {n = 0} ^ {N-1} \ alpha \ text {*} (k, n) \ cdot x (m + nNc)

This expression is similar to the one that facilitates and (k, m) from the coefficients h * (iTe) and the samples r (mTe), but leads to a simpler material realization than that of the general case. As in the general case, the samples and (k, m) can be obtained by sliding correlation: samples x (m) transit in a shift register with the rhythm He; to each position nNc, multiple of Nc, the register will be Associates the α * chip (k, n) used to calculate the product α * (k, n)? (m + nNc); in each period Te, you It supplies a new sample and (k, m). It should be noted that the Length of the shift register is M = N \ cdotNc, while the number of products is only N. The chips α * (k, n) must be renewed under the same conditions as the coefficients h * (k, iTe) of the general case.

From the same series of samples x (m) from filtration adapted to the shape of the chip, the widening in follow-up supplies the series of samples:

z (k, m) = \ sum \ limits_ {n = 0} ^ {N-1} \ alpha \ text {*} (k, n) \ cdot x (m + nNc)

evolving the range k according to indications provided by the tracking device during the follow-up phase. As in the general case, the complexity of the widening in follow-up only depends on the Ne length of the window.

       \ global \ parskip1.000000 \ baselineskip
    

Figures 4 and 5 illustrate the procedure of the invention in simplified situations. They correspond to the case of a transmission with an editing function h (k, t) that depends of the range k of the symbol, as explained above. For simplicity, it is assumed that there were only two paths of propagation and that the power calculations refer to a single symbol (the powers P (p, q) and Pa (p, q) are then identical). For this example, an operation with eight is assumed samples per symbol (M = 8) and a window with a length of 5 samples (Ne = 5).

In Figure 4, line (a) represents the sample powers, line (b) successive positions of the sliding window of NeTe length, line (c) power P (p, q) in each window, line (d) the range m of Track and line (e) the editing range.

Figure 4 shows how information is passed from the sliding sliding window to the tracking window. The scan spread correlator uses the reference chip sequence that corresponds to an arbitrarily selected range k_ {0}, but fixed throughout the transmission. The sequence of the samples r (mTe) parades in the correlator; therefore, all possible relative positions of the reference sequence and the received sequence are successively examined and observed on the evolution of the power | y (k_ {0}, m) | 2}
elementary two maximums corresponding to the coincidence with the propagation paths. The power P (p, q) obtained by the sum of the powers | y (k_ {0, m) | 2} is constant and low outside the area of these propagation paths; its value increases when the first route is taken into account, passes through a plateau, then increases again when the second route is taken into account; then it decreases by plateaus when the sliding window leaves the first path and then the second. Each time the power P (p, q) increases, a symbol range k and a sample range m are transmitted to the tracking device (downward arrows between lines (c) and (d)). The range k is none other than k_ {0}, since it is the range of the reference sequence used in scanning for which the power increase has just occurred. When the samples of m = 0 am = Ne-1 are numbered in the tracking window, the transmitted m range is equal to Ne-1 since it can be assumed in a first approximation that the window closes shortly after the passage of power through a maximum Taking into account the first maximum power triggers a synchronization that is not good; the passage of the second maximum produces the definitive synchronization. Outside the transfers, the tracking range m is increased by 1, module M (line (d)) and the range k is increased each time the range m passes through the value 0 (line (e)).

In Figure 5, line (a) represents the power of the samples under exploration, line (b) the range k, the line (c) range m, line (d) the place of the window of tracking, line (e) the power of the samples of tracking, line (f) deviation from position desired.

After the appearance of the two maximums corresponding to the propagation paths (line (a)), the power
| y (k_ {0}, m) | ^ {2} no longer shows significant maximums in the periods of symbols that follow, until the editing function used in the broadcast is again the one corresponding to the range k_ {0} . On the contrary, the power | z (k_ {0}, m) | 2 (line (e)) is obtained from a reference sequence in which the range k evolves in each symbol period of the same way as the emission range; therefore the two correlation maxima are observed in each symbol period. From the distribution of the powers | z (k_ {0}, m) | 2 in the window, the tracking device calculates an ideal position, for example the barycenter of the powers. The deviation between this position and the center of the window is used to control the rhythm generator. Thus, the deviation represented in Figure 5 (line (f)) slows down the rhythm generator in order for the center of the window to move towards the barycenter of the powers.

It should be noted that, in most of the cases, the samples are complex quantities with a real part and An imaginary part. Thus, the sample series r (mTe) is composed of two real series r_ {I} (mTe) and r_ {Q} (mTe): r (mTe) = r_ {I} (mTe) + j \ cdotr_ {Q} (mTe). The family of the coefficients h * (iTe) is composed of two real families h_ {I} (iTe) and h_ {Q} (iTe): h * (iTe) = h_ {I} (iTe) -j \ cdoth_ {Q} (iTe). Then the real and_ {I} (m) and imaginary parts are provided y_ {Q} (m) of the sample and (m) for the expressions:

y_ {I} (m) = \ sum \ limits_ {i = i_ {1}} ^ {i_ {2}} \ {h_ {I} (iTe) \ cdot r_ {I} [(i + m) \ cdot Te] + h_ {Q} (iTe) \ cdot r_ {Q} [(i + m) \ cdot Tea]\}

y_ {Q} (m) = \ sum \ limits_ {i = i_ {1}} ^ {i_ {2}} \ {h_ {I} (iTe) \ cdot r_ {Q} [(i + m) \ cdot Te] - h_ {Q} (iTe) \ cdot r_ {I} [(i + m) \ cdot Tea]\}

in which the calculations are performed About real numbers. Likewise, the power of the sample and (m) by the sum of the powers of the parts real e imaginary:

[y (m)] 2 = [y_ {I} (m)] 2 + [y_ {Q} (m)] 2

The development of the expressions they make intervene the real \ alpha_ {I} (n) and imaginary parts [alpha] {Q} (n) of the [alpha] (n) chip, or of the real parts z_ {I} (m) and imaginary z_ {Q} (m) of the Sample z (m) is similar and immediate. In certain cases particular, any of these elements can be reduced to one series of real values, simplifying the calculations.

       \ newpage
    

The realization of the adapted filtrations is has submitted in the form of displacement records in which samples and records of coefficients or chips pass through. Other equivalent methods are possible; in particular, they can used generators of coefficients h * (iTe) or chips α (n) that provide the coefficient or chip corresponding to the current r (mTe) or x (m) sample, and get the sample y (m) or z (m) by accumulation of products of type h * (iTe) \ cdotr [(i + m) \ cdotTe] or? * (n) \ cdotx (m + nNc).

By way of illustration of the above description, the case of an AMDC transmission with widening can be described by direct sequence and phase shift modulation in 2 states (MDP2). The symbols a (k) are binary symbols that take the values +1 and -1. Binary elements are called often 0 and 1, but in this case it is the values +1 and -1, which they are associated respectively with these binary elements 0 and 1 in view of modulation. Each symbol a (k) is widened by a sequence of N chips? (k, n), for from n = 0 to N-1, depending on the range k but independent of symbol value a (k). Each chip α (k, n) takes one of the values +1 or -1. The chip family α (k, n) it comprises K successive sequences of N chips, numbered from k = 0 to k = K-1. Therefore it is reused for all K Symbols The g (t) chip editing function is a function That takes real values. It is selected so that the frequency maximum contained in its spectrum is less than the Hc rhythm of the chip, what is usually the case. It is enough, for example, to take a function that respects Nyquist's condition with respect to rhythm of the chip and whose spectrum has the classic form of a "cosine elevation "of rounding factor less than 1. Taking into account this property, the received signal r (t) of baseband can be sampled with the rhythm 2 \ cdotHc, that is, 2 samples per chip, no loss of information (provided that the filtration analog preceding sampling suppress higher frequencies to Hc).

From the sample series r (mTe) of baseband, the filtration adapted to the shape of the chip elaborates a series of samples x (m) in which the actual parts x_ {I} (m) and imaginary x_ {Q} (m) are provided by Simple expressions that are made on real numbers:

x_ {I} (m) = \ sum \ limits_ {i} g (iTe) \ cdot r_ {I} [(i + m) \ cdot Tea]

x_ {Q} (m) = \ sum \ limits_ {i} g (iTe) \ cdot r_ {Q} [(i + m) \ cdot Tea]

The real y_ {I} (k, m) and imaginary parts y_ {Q} (k, m) of the sequence y (k, m) supplied by the spreading device are facilitated by the expressions following, that are realized on real numbers:

y_ {I} (k, m) = \ sum \ limits_ {n + 0} ^ N-1} \ alpha (k, n) \ cdot x_ {I} (m + 2n)

y_ {Q} (k, m) = \ sum \ limits_ {n + 0} N-1} \ alpha (k, n) \ cdot x_ {Q} (m + 2n)

Thus, it seems that, for a sampling range m given, the calculation of y_ {I} and y_ {Q} only uses a sample of every two of the series x (m). Similar conclusions are obtained for the sample series z (m).

The adapted filtration and synchronization scheme of Figure 8 is deduced. The signal 51A received sampled is composed of two series of real samples r_ {I} (mTe) and r_ {Q} (mTe) that are applied respectively to two filters 50I and 50Q separated from the same transfer function g (-t) that perform the filtration adapted to the chip. These filters constitute a common part of the filtration 52 adapted in scanning and of the filtration 53 adapted in monitoring. The series of samples x_ {I} (m) and x_ {Q} (m) from these filters are applied simultaneously to the scaling devices under exploration and monitoring. Exploration de-widening is comprised of a chip generator 520 and two de-widening circuits 52I and 52Q themselves that function respectively on the sample series x_ {I} (m) and x_ {Q} (m) to supply the series of samples y_ {I} (m) and y_ {Q} (m). Similarly, the de-widening in follow-up is composed of a chip generator 530 and two circuits 53I and 53Q of de-widening operating on the series x_ {I} (m) and x_ {Q} (m) to supply the sample series z_ {I} (m) and
z_ {Q} (m). The scanning device 72 operates according to the general indications given above: calculates the elementary powers [y_ {I} (m)] 2 + [y_ {Q} (m)] 2, then the power in one sliding window and looks for the relative position of the chip sequence with respect to the samples received that lead to maximum power in the window. Similarly, the tracking device 73 works according to the general case: calculates the elementary powers
[z_ {I} (m)] 2 + [z_ {Q} (m)] 2, the power in a tracking window, the feedback signal d of the rhythm generator 71; Performs the required checks and supplies the Hs rhythm of the regenerated symbol.

The filter (50I or 50Q) adapted to the shape of the Chip is a classic digital filter. It can be done, for example, by means of a shift register in which they transit in the rhythm I have samples r_ {I} (mTe) or r_ {Q} (mTe); to each position of the register is associated with the coefficient g (iTe) by which the current sample of that position is multiplied; the sum of the products obtained constitutes in a given moment the sample x_ {I} (m) or x_ {Q} (m) current.

The chip generator (520 or 530) is composed, for example, of a table in which the chips are ordered  successive and by a counter register containing the current range of the chip to be used. This counter is incremented by one unit every two sampling periods and the chip is supplied corresponding to the stripping circuits. If the law of sequence formation of the K \ cdotN chips is sufficiently simple, for example if the chip sequence can be deduced from any part of the given length of the sequence is possible Other chip generation procedures. In general, the transfer of the sliding window position of scanning to the tracking window consists of transmitting the range of the 520 generator chip counter in scan or a part of the corresponding chip sequence, at the time in the that the power obtained in the sliding window passes through a maximum.

The scanning de-widening circuit (52I or 52Q) is composed, for example, of a sample register, a chip register, a transit register of the chips and summation devices. The sample register is a shift register in which the samples x_ {I} (m) or x_ {Q} (m) travel in the He rhythm. The chip transit record is a shift register in which the chips from the 520 generator run at the chip rate (that is, in this case at half the sampling rate I). Every 2N sampling periods (or every 2N \ cdotNs periods, if Ns are treated simultaneously in a sample register containing 2N \ cdotNs samples), the contents of the chip transit record are transferred to the chip register. The content of the chip register remains immutable between two transfers. One position in two of the sample register is associated (both even range positions and odd range positions can be selected) to one position in the chip register. According to the value +1
or -1 of the chips, the sum device corresponding to a symbol extracted for each chip, the current associated sample or its opposite and performs the sum of the extracted values; the result constitutes a sample y_ {I} (m) or y_ {Q} (m). As indicated above, the chip generator 520 should be conveniently reset every time 2N • relative positions are examined.

The de-widening circuit (52I or 52Q) in Scanning can also be done with the help of a record of displacement in which the chips, in the rhythm of the chip, travel from generator 520. At each position in this register, They are associated with two accumulator devices. An accumulator device performs, in the rhythm of the chip, the sum, during a period of symbol, of the current sample x_ {I} (m) or x_ {Q} (m) of even range or the opposite value of this sample depending on whether the present chip is worth +1 or -1. The second device performs the same operation on odd range samples. So, if the record of displacement comprises N positions, are obtained successively 2N samples, corresponding to 2N relative positions consecutive; if the shift register comprises N \ cdotNs positions, Ns corresponding samples are obtained simultaneously  to relative positions separated by 2N and, after 2N periods of sampling, 2N \ cdotNs relative positions have been examined consecutive.

The de-widening circuit (53I or 53Q) in follow-up can be done in a manner similar to that of desenschamiento in exploration, but its extension is very reduced For example, if the second procedure is selected embodiment, mentioned above for the widening in scan, only Ne accumulation devices are needed, in which Ne is the number of samples contained in the window.

Figures 6 and 7 illustrate these techniques of de-widening, or by the method of records of displacement (figure 6), or by the accumulation method (figure 7). These figures always refer to the case of a AMDC transmission with MDP2 modulation, for received signals sampled at the rate of two samples per chip. They facilitate a representation of the principle of the widening circuits that can be used (circuits represented by 52I and 52Q for exploration, 53I and 53Q for monitoring). Only represented the circuit of track I (being identical to that of track Q). He α (k, n) reference chip generator is not It represents. For simplicity of the drawing, it is limited to the case in which the correlation is made on a single symbol (that is, on 2N samples, N being the broadening factor) and in which the window length is less than the duration of a symbol (Ne <2N). Control signals are obtained from the He rhythm Sample by means of a 2N module counter. The even state odd counter plays in the chip beat while any particular range of the counter plays in the rhythm of symbol.

Figure 6 thus shows a module counter 2N with reference number 100 with an entry that receives the rhythm I sampling and two outputs corresponding to an even state (rhythm of chip) and at a state 0 (symbol rate), a record 102 of samples in 2N positions that receive a shift signal that is He and, as input, the sample x_ {I}, a chip register 104 which receives a read signal that is the 0 output state of the counter 100, a traffic register 106 controlled by the signal "even state" and the chips a (k, n) and multipliers 110 that multiply the samples by the chips, being connected the outputs of these multipliers to an adder 108, in which the output provides the signal y_ {I}.

The sample sequence runs in a manner permanent in the register 102 of samples. The chip sequence is enter in the transit register 106, according to conditions that They depend on the use (exploration or monitoring). It is transferred by blocks to each symbol period in chip register 104. The correlation is done every two positions of the register of samples.

On the contrary, Figure 7 shows how the widening can be obtained by accumulating the partial correlations of the current sample received x_ {I} and a sliding part of the sequence of chips α (k, n). The represented circuit comprises a 2N module counter with reference number 120 with an input that receives the sampling rate He, and three outputs that provide a signal "even state", "odd state" and "count range", a chip register 122 that receives the "even state" signal and, as input, the chips
α (k, n), multipliers 124 with two inputs, one connected to the output of a cell in register 122 and the other to a line 125 that receives the current sample x_ {I}, sets 130, 140, etc. Accumulation comprising a door 131 (141) controlled by the signal "even state"("oddstate") and by the output of the multiplier 124, an initialization circuit 132 (142) that receives the signal "range = 0"("range = 1"), an adder-accumulator 133 (143) connected to the gate 131 (141), the output of this adder being reset on the initialization circuit 132 (142), a controlled reading circuit 134 (144) by the signal "rank 0"("rank1") and which supplies the samples y_ {I} (k, 0) dismantled, (or respectively y_ {I} (k, 1)).

Claims (14)

1. Procedure for synchronizing the rhythm of a digital signal, in which an analog signal (r (t)) is sampled with a certain sampling period (Te) from the transmission of a modulated signal (b (t)) with With the aid of an editing function (20), filtration (52, 53) adapted from the samples (r (mTe)) is performed, this filtration being adapted to the editing function used by the modulation and leading to samples (and (m) 52A, z (m) 53A) of correlation, the power of each correlation sample (and (m)) is calculated, a sliding window of length is defined Ne times the sampling period (Te), ie NeTe , and that begins in a certain range (m_ {0}), sweeping this sliding window a certain interval of sampling moments, characterized in that: - for each sliding window, one is calculated global power equal to the sum (Pa) of the powers of the correlation samples (and (m)) located in this window to a symbol or for a certain number of symbols; - the window for which the sum of sample powers is maximum; - synchronization is then defined by the position of the window in which the sum of powers of the samples is maximum, and by the range of each correlation sample in The inside of this window. 2. Method according to claim 1, in which two types of operations are implemented and proceeds in two modes: a) in a first type of operations, called exploration operations, all the following are examined successively Possible positions of the sliding window, and for each position, the overall power (Pa) of the samples (y (m)) of correlation contained in the sliding window, the window for which the overall power is higher from the cycle start to the current position, and the value is memorized of this superior power (Pam), b) in a second type of operations, called follow-up operations, only the correlation samples (z (m)) in which the range is in a window called tracking window, is calculated by a part the overall power (Pb) of these samples, on the other hand a signal (d) that allows feedback to the center of this window about the average position of the powers of the samples that contains, c) in a first mode of operation, called acquisition mode,

-

by a part, every time a sliding window appears a overall power (Pa) greater than the last memorized power (Pam) from the beginning of the scan cycle to the current position, the current position of the sliding window and a procedure of check,

-

by another part, when the scan cycle is finished, it goes to a called tracking mode,

d) in a second mode of operation, called of follow up,

-

by one part, the transfer mechanism of the position of the sliding window to the tracking window,

-

by otherwise, when permanent verification fails, it returns to the acquisition mode

3. Method according to claim 2, in which a relative threshold λ of power is selected between 0 and 1, the product of this threshold is calculated by the highest memorized power and it is verified that the overall power (Pb) obtained in the tracking window is statistically superior to the value of the product. 4. Method according to claim 3, in the one that performs a first verification, initial call, that it consists in examining whether, between the Na successive values of the global power (Pb), at least Nb among them are greater than threshold product by the higher power, selecting the Na and Nb numbers in the order of some units, and if applicable, it goes to a second verification, permanent call, that consists in comparing the global power (Pb) with an amount \ muPo, where \ mu is a threshold slightly below 1 and in which Po designates the expected average power (Pb) in the state synchronized, comparing over a large number of successive values of the global power (Pb). 5. Method according to claim 4, in which, in acquisition mode, the appearance, in a window sliding, of a global power (Pa) higher than the last memorized power (Pam), re-launch a new verification initial. 6. Method according to claim 2, in which the global powers (Pa, Pb) calculated in the windows sliders and in the tracking window result in the sum of the powers obtained for various symbols.

         \ newpage
      

7. Method according to claim 2, in which is associated with a weight c (i) to each sampling range (i) inside the tracking window, this weight being in function of the position of this range with respect to the center of the window, it is determined where the average position of the powers of the weighted samples with respect to the center of the window and the signal sampling moments are corrected analog (r (t)) to feed back the center of the window about this middle position. 8. Method according to claim 7, in that the weights c (i) are of the form: c (i) = i- (Ne-1) / 2 in which the amount (Ne-1) / 2 characterizes the center of the window. 9. Method according to claim 7, in which weights (c (i)) are taken equal to +1 in half top of the window and equal to -1 in the bottom half of the window. 10. Procedure according to any one of the claims 1 to 9, wherein the sampling period (Te) of the analog signal (r (t)) is a fraction of the period (Ts) of the symbols. 11. Procedure according to any one of the claims 1 to 10, wherein the analog signal (r (t)) corresponds to a signal of the multiple access type by division of code corresponding to spreading sequences formed by numbers (α (n)) called chips, to which it applies an edition, procedure in which the adapted filtration performed on the analog signal (r (t)) comprises a first filtration adapted to the shape of the chips and that supplies the first samples (x (m)) and a second filtration adapted to the widening sequences and applied to the first samples, supplying this second filtration adapted second samples (y (m), z (m)) from those that calculate the powers of the samples and the powers global respectively located in the sliding windows and of tracing. 12. Method according to claim 11, in which the sampling of the analog signal is performed at a frequency equal to k times the frequency of the chips that form the widening sequences, then using the second adapted filtration, at a given moment, a sample of each k of The first samples. 13. Method according to claim 12, in which k is equal to 2, thus sampling the signal analog at a frequency twice the frequency of the chips that form the broadening sequences, using the second adapted filtration, at any given time, one in two samples. 14. Procedure according to any one of the claims 1 to 13, wherein the analog signal (r (t)) it is a complex signal with a real component and a component imaginary, and in which each sample is also complex with a real component and an imaginary component, running the adapted filtration on the real and imaginary components, calculating the powers of the samples and the global powers realizing the sum of the powers of the real components and Imaginary samples.